Scene inversion impairs activation of scene-object semantic bias.

Abstract

Inferences on semantic scene content enable the visual system to efficiently parse complex naturalistic scenes. In the early stages of scene perception, conceptual scene gist activates semantic expectations and relevant scene-object associations that bias the viewer to affirm the presence of semantically consistent objects and to reject the presence of semantically inconsistent objects within the overall scene context. This contextual scene-object bias is crucial to various visual cognitive functions, from object recognition to attentional guidance. Despite its top-down influence, however, few studies have directly investigated the factors that drive the activation of scene-object bias. Here, we tested the role of global scene semantics in the activation of scene-object bias by comparing response bias on briefly presented upright versus inverted scenes. We found that scene inversion not only significantly delayed overall bias activation, but also differentially impacted superordinate-level versus basic-level bias. Hierarchical differences were also highlighted by earlier and more robust activation of superordinate-level bias than basic-level bias. Taken together, the present work provides important characterization to gist-activated scene-object bias, which is an essential and influential top-down mechanism in scene perception and visual cognition.